The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze various untrusted data sources to facilitate research and clarify user requirements. This creates a surface where malicious instructions embedded in project artifacts or external data could attempt to influence the agent's behavior.
Ingestion points: SKILL.md (Step 2: Research before asking) specifies that the agent should inspect available code, docs, tickets, plans, tests, logs, schemas, configuration, runtime evidence, prior discussion, and current external facts.
Boundary markers: There are no instructions in the skill to use delimiters or explicit 'ignore instructions' warnings when processing these external data sources.
Capability inventory: The skill's primary capabilities involve generating text-based briefs and questions for the user. It does not include direct commands for filesystem modification or network exfiltration within its own instructions, though it may utilize platform tools for research.
Sanitization: The instructions do not define any sanitization, filtering, or validation processes for the ingested research data.

grill-me